Discover Awesome MCP Servers

Knowledge Graph Builder

Transforms text or web content into structured knowledge graphs using local AI models with MCP integration for persistent storage in Neo4j and Qdrant.

Google Calendar MCP Server

Servidor de Protocolo de Contexto del Modelo (MCP) que se integra con la API de Google Calendar.

AI Agent with MCP

Okay, here's a basic outline and code snippets to help you create your first MCP (Model Context Protocol) server in a Playground environment. Keep in mind that MCP is a relatively new and evolving protocol, so the specific libraries and implementations might change. This example focuses on a simplified, conceptual approach. **Conceptual Overview** 1. **Choose a Language/Framework:** Python is a good choice for rapid prototyping and has libraries suitable for networking and data serialization. 2. **Define Your Model:** Decide what kind of model you want to serve. For a simple example, let's imagine a model that performs basic arithmetic (addition). 3. **Implement the MCP Server:** * Listen for incoming connections. * Receive MCP requests. * Parse the requests. * Execute the model (in our case, addition). * Format the response according to MCP. * Send the response. 4. **Implement a Simple MCP Client (for testing):** * Create a client to send requests to your server. * Receive and parse the responses. **Simplified Python Example (using `socket` and basic JSON)** ```python # server.py (This would run in your Playground) import socket import json HOST = '127.0.0.1' # Standard loopback interface address (localhost) PORT = 65432 # Port to listen on (non-privileged ports are > 1023) def handle_request(data): """ Simulates a simple model that performs addition. Assumes the data is a JSON string with 'a' and 'b' keys. """ try: request = json.loads(data.decode('utf-8')) a = request.get('a') b = request.get('b') if a is None or b is None: return json.dumps({"error": "Missing 'a' or 'b' parameter"}).encode('utf-8') try: result = a + b response = {"result": result} return json.dumps(response).encode('utf-8') except TypeError: return json.dumps({"error": "Invalid 'a' or 'b' value (must be numbers)"}).encode('utf-8') except json.JSONDecodeError: return json.dumps({"error": "Invalid JSON"}).encode('utf-8') with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s: s.bind((HOST, PORT)) s.listen() print(f"Server listening on {HOST}:{PORT}") conn, addr = s.accept() with conn: print(f"Connected by {addr}") while True: data = conn.recv(1024) # Receive up to 1024 bytes if not data: break # Client disconnected response = handle_request(data) conn.sendall(response) # Send the response back to the client ``` ```python # client.py (This would run in a separate Playground or terminal) import socket import json HOST = '127.0.0.1' # The server's hostname or IP address PORT = 65432 # The port used by the server with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s: s.connect((HOST, PORT)) # Example request (MCP-like, but simplified) request_data = {"a": 5, "b": 3} request_json = json.dumps(request_data).encode('utf-8') s.sendall(request_json) data = s.recv(1024) print('Received:', repr(data.decode('utf-8'))) ``` **Explanation and Key Points:** * **`server.py`:** * Sets up a basic TCP socket server. * `handle_request()`: This is where your "model" logic goes. In this example, it's a simple addition function. It receives JSON data, parses it, performs the addition, and returns a JSON response. Error handling is included. * The server listens for connections, accepts a connection, and then enters a loop to receive data, process it, and send a response. * **`client.py`:** * Creates a TCP socket client. * Connects to the server. * Constructs a JSON request (representing an MCP-like request). * Sends the request to the server. * Receives the response and prints it. * **JSON for Serialization:** JSON is used for encoding and decoding the requests and responses. This is a common and relatively simple way to handle data serialization. * **Error Handling:** Basic error handling is included in the `handle_request` function to catch invalid JSON, missing parameters, and type errors. * **Simplified MCP:** This is *not* a full MCP implementation. It's a simplified example to illustrate the basic concepts. A real MCP implementation would involve more complex message structures, metadata, and potentially other protocols for data transfer. * **Playground Considerations:** Make sure your Playground environment allows network connections. Some online Playgrounds might have restrictions. If you have issues, try running the server and client on your local machine. * **Running the Code:** 1. Run `server.py` in one Playground or terminal window. 2. Run `client.py` in another Playground or terminal window. 3. The client should connect to the server, send the request, and print the response. **To make this more like a real MCP server, you would need to:** * **Define a formal MCP message structure:** MCP has specific requirements for the format of requests and responses, including metadata and data encoding. You'd need to adhere to those specifications. * **Implement a more sophisticated model:** Replace the simple addition with a more complex machine learning model. You might use libraries like TensorFlow, PyTorch, or scikit-learn. * **Handle different data types:** MCP needs to support various data types (images, text, etc.). You'd need to implement appropriate serialization and deserialization methods. * **Add authentication and authorization:** Secure your server to prevent unauthorized access. * **Consider performance:** Optimize your code for speed and efficiency, especially if you're serving a high volume of requests. **Important Considerations for Playgrounds:** * **Network Access:** Many online Playgrounds have limited or no network access. If you can't get the server and client to connect, it's likely a network restriction. Try running the code locally on your machine. * **Dependencies:** Make sure your Playground environment has the necessary libraries installed (e.g., `json`). If not, you might need to install them using `pip` or a similar package manager. * **File System Access:** Some Playgrounds might restrict file system access. If you need to load model files, you might need to find alternative ways to store and access them (e.g., using cloud storage). This example provides a starting point. You'll need to research the specific MCP specifications and adapt the code to your particular model and requirements. Remember to consult the official MCP documentation and any relevant libraries for more detailed information.

Math-Physics-ML MCP System

Provides GPU-accelerated scientific computing capabilities including symbolic mathematics, quantum wave mechanics simulations, molecular dynamics, and neural network training through four specialized MCP servers.

TinyDB MCP Server

Provides long-term memory for chatbots through a TinyDB-backed MCP server. Enables storing, searching, and managing JSON records with schema validation without requiring external databases.

Semantic D1 MCP

Enables AI-assisted analysis and optimization of Cloudflare D1 databases through comprehensive schema introspection, relationship mapping, validation, and optimization recommendations. Demonstrates semantic intent patterns with hexagonal architecture for maintainable database development workflows.

MCP Salesforce Revenue Cloud

Provides AI assistants with direct access to Salesforce Revenue Cloud data and operations, enabling retrieval of products, price books, quotes, orders, and execution of custom SOQL queries through natural language.

DeepSeek Thinking with Claude 3.5 Sonnet

Facilita procesos de razonamiento en dos etapas utilizando DeepSeek para un análisis detallado y admite múltiples modelos de respuesta como Claude 3.5 Sonnet y OpenRouter, manteniendo el contexto de la conversación y mejorando las interacciones impulsadas por la IA.

MCP Agent Platform

Un sistema de interacción persona-ordenador multiagente que permite la interacción natural a través de capacidades integradas de reconocimiento visual, reconocimiento de voz y síntesis de voz.

MCP Debugger Server

Provides comprehensive debugging capabilities for Node.js and TypeScript applications with 25+ specialized tools, including breakpoint management, variable inspection, hang detection, CPU/memory profiling, and test framework integration.

DatahubMCP

Enables Claude to query MySQL databases and access Google Workspace (Sheets, Forms, Drive) for education program management, with built-in templates for data analysis and report generation.

YApi MCP Server

Enables reading and searching API documentation from YApi instances, allowing AI models to access interface definitions, project API lists, and search through API endpoints using YApi URLs or project IDs.

MCP Express Server

A TypeScript-based MCP server template using Express.js and Server-Sent Events, providing example tools for echoing messages, performing calculations, and retrieving server time.

Sirr MCP Server

Provides Claude Code direct access to a Sirr secret vault for reading, pushing, listing, and deleting secrets with expiry constraints. It enables natural language secret management while keeping credentials secure through metadata-only listing and controlled value retrieval.

Affinity MCP Server

Enables AI assistants to control the Affinity creative suite on macOS through natural language, allowing for automated design tasks, UI interaction, and file operations. It leverages AppleScript and System Events to bridge AI commands with professional creative software.

APK Security Guard MCP Suite

Provides a one-stop automated solution for Android APK security analysis by integrating tools like JEB, JADX, APKTOOL, FlowDroid, and MobSF into unified MCP standard API interfaces.

n8n MCP Server

Enables full workflow automation management in n8n through 40+ tools covering workflows, executions, credentials, tags, variables, projects, users, and source control operations.

mcp-server-bitbucket

MCP server with 58 tools for Bitbucket API operations. Manage repositories, pull requests, pipelines, branches, commits, deployments, webhooks, and more.

Palo Alto Networks MCP Server Suite

Enables comprehensive management of Palo Alto Networks firewalls through a modular suite of servers for security policies, network objects, device operations, and system configuration.

LlamaCloud MCP Server

Un servidor MCP local que se integra con Claude Desktop, habilitando capacidades RAG para proporcionar a Claude información privada y actualizada desde índices LlamaCloud personalizados.

Atlassian MCP

Plugin de código gestionado para Cursor IDE que proporciona integración con productos de Atlassian (JIRA, Confluence, BitBucket), permitiendo a los desarrolladores buscar tareas, crear nuevos problemas, ver documentación y gestionar repositorios de código directamente desde el IDE.

Ensembl MCP Server

A Model Context Protocol server providing LLMs with access to the Ensembl genomics database, enabling AI assistants to query gene information, sequences, variants, and other genomic data across multiple species.

Claude File Management Server

Enables Claude AI to perform comprehensive file operations including reading, writing, notes management, PDF processing, and image handling with thread-safe operations. Provides a complete file management system with enhanced features like document processing and automatic directory creation.

Scrapy MCP Server

A powerful web scraping MCP server built on Scrapy and FastMCP that supports multiple scraping methods (HTTP, Scrapy, browser automation), anti-detection techniques, form handling, and concurrent crawling. Designed for commercial environments with enterprise-grade features like intelligent retry mechanisms, performance monitoring, and configurable data extraction.

Pentest MCP

Un servidor de Protocolo de Contexto de Modelo que integra herramientas esenciales de pruebas de penetración (Nmap, Gobuster, Nikto, John the Ripper) en una interfaz unificada de lenguaje natural, permitiendo a los profesionales de seguridad ejecutar y encadenar múltiples herramientas a través de comandos conversacionales.

File System MCP Server

Provides secure file read and write operations within a sandboxed directory, allowing AI assistants to safely create, modify, and access files without risk of accessing the broader file system.

Viterbit MCP Server

Enables interaction with Viterbit recruitment API for managing candidates, jobs, and applications. Supports searching, updating candidate data, handling job applications, and advanced filtering with subscription and activity status tracking.

YARR Media Stack MCP Server

Un servidor de Protocolo de Contexto de Modelo integral que conecta LLM con servicios de medios autoalojados, permitiendo el control en lenguaje natural de programas de televisión, películas, descargas y notificaciones, al tiempo que mantiene el acceso a la API tradicional.